85% of your people have the skills to use AI. About 25% actually use it at work. That 60-point gap isn't a training problem, a talent problem, or a tooling problem. It's a last-mile problem — and almost every organization I walk into has it.
The IBM Institute for Business Value put real numbers on this in its 2026 CEO study. 76% of CEOs now have a Chief AI Officer or equivalent, up from 26% two years earlier. The strategy is in place. The budget is approved. The licenses are bought. And then the workflow doesn't change, so people quietly default back to the way they did the work last year.
I've watched this pattern across 300+ engagements. It's remarkably consistent. The good news is that consistency cuts both ways — if the failure is this predictable, so is the fix.
The gap is in the last mile, not the rollout
Most AI initiatives die in the same place: the distance between "we have the tool" and "this is how the work gets done now." That last mile is where adoption actually happens, and it's the part almost nobody owns.
Think about what a typical rollout looks like. Leadership picks a tool. IT provisions seats. There's a kickoff, maybe a lunch-and-learn, a Slack channel, a deck. Everyone nods. Then Monday comes, the deals still need to close, the tickets still need answers, and the AI step is one more tab nobody has to open. So they don't.
The skills were never the bottleneck. IBM's own number tells you that — 85% already have what they need. The bottleneck is that the new behavior is optional, and optional loses to habit every single time.
The four-part failure pattern
When I diagnose a stalled rollout, I'm looking for four specific things. They show up together so often I treat them as a checklist.
A pilot with no owner. The initiative belongs to "the team" or "the AI working group," which means it belongs to no one. There's no single person whose week gets worse if adoption stays flat. Diffuse ownership is how a pilot becomes a screensaver.
A tool bolted onto an unchanged process. The workflow people actually run — the CRM stage, the support queue, the proposal draft — stays exactly as it was. The AI lives beside it, in a separate window, as an extra step. You haven't changed the process. You've added homework to it.
No trigger that forces the moment of use. Nothing in the day says "now you use the AI." There's no required field, no stage gate, no handoff that routes through it. Use depends entirely on someone remembering and choosing to, under deadline pressure, when the old way is right there and faster to reach for.
No measurement. Nobody can tell you what percentage of eligible work actually ran through the tool last week. Without that number, "adoption" is a feeling, and feelings drift toward optimism. You can't fix what you're not counting.
Hit three of these four and your 25% is locked in. I've never seen it otherwise.
Optional behavior loses to habit every time. The AI step has to be the easy path, not the extra one.
The fix is embedding, not encouraging
Encouragement doesn't move adoption numbers. Architecture does. This is the core of the AI Adoption Framework™ I use with clients, and it comes down to three moves you can start this Monday.
Embed the AI inside a workflow people already run. Don't ask anyone to go somewhere new. Put the AI step where the work already lives — inside the CRM record, the ticket, the document they were going to open anyway. The win isn't a better tool. It's removing the decision to use it. If using the AI requires opening a second app, you've already lost most of your 60 points.
Make the AI step the path of least resistance. People take the easy road, so make the AI road the easy one. The draft should already be written when they arrive. The summary should already be at the top of the record. The next-best-action should be sitting in the field, waiting to be accepted or edited. When the AI output is the starting point instead of an errand, "use it" stops being a choice and becomes the default.
Pick one process, instrument it, measure weekly. Not the whole company — one process. Lead follow-up, or ticket triage, or proposal drafting. Define what "used" means in concrete terms, put a number on it, and look at that number every week. One process moving from 25% to 70% teaches the organization more than ten pilots that never got measured. And it gives you the one thing strategy decks never do: proof the new way actually sticks.
Why this is worth getting right
The 60-point gap isn't an efficiency footnote. It's the difference between an AI budget that compounds and one that quietly becomes shelfware your CFO remembers next year. The companies pulling ahead right now aren't the ones with better models or bigger budgets. They're the ones who closed the last mile while everyone else was still running kickoffs.
You already bought the tools. Your people already have the skills. What's missing is the architecture that makes use the default instead of the exception — and that's a design problem, which means it's solvable.
If you want the weekly version of this thinking — the specific patterns I see across client work, and the fixes that actually move the number — that's what my newsletter is for. No theory, just what works.
